Mistral Shieldstral safety model posts a reported tie with OpenAI’s 20B guard
Mistral released its 3B Shieldstral safety model under Apache 2.0, reporting an 84.9% text F1 tie with a 20B OpenAI guard model.
By Colin Brandt 3 days agoSection
AI
Keyv npm provenance attack shows the limit of a valid security attestation
A malicious Keyv release received genuine SLSA provenance because it passed through the project’s trusted GitHub Actions workflow.
By Renata Fuchs 3 days ago
ResumeUp.AI launches conversational resume builder after claimed 500,000-user mark
ResumeUp.AI says it now serves more than 500,000 users as it replaces form-led resume creation with a real-time AI chat workflow.
By Colin Brandt 3 days agoAI coding agent cost controls at Replit, Kilo Code and Symbotic
Replit, Kilo Code and Symbotic outlined model routing, spending tiers and review controls for AI coding agents.
By Colin Brandt 4 days ago
White House Chinese AI restrictions remain unannounced after industry split
The administration weighed limits on Chinese open-weight AI, but no blanket ban was announced as startups and large tech firms obj...
By Colin Brandt 4 days agoAsana AI Teammates use shared memory with access-control claims
Asana says its AI Teammates retain workflow context across teams, but buyers still need to test how permissions isolate confidenti...
By Colin Brandt 4 days agoQwen3.8-Max OSWorld benchmark claim puts Alibaba ahead of named rivals
Alibaba says Qwen3.8-Max scored 86.1 on OSWorld-Verified, but the computer-use result and long-running demos await independent tes...
By Wei-Lin Zhao 4 days agoNTT DATA AIVista makes its case for the last mile of enterprise AI
At VB Transform 2026, AIVista outlined how it says proprietary context, guardrails and operations can move agents into regulated w...
By Wei-Lin Zhao 5 days ago
MiniMax H3 video ranking leads editing, but open-model claim needs qualification
MiniMax H3 took first place in Artificial Analysis video editing, while its later weight release came with material license and pr...
By Wei-Lin Zhao 5 days ago
Alibaba’s Qwen3.8-Max open weights are due next week
Alibaba has opened hosted access to Qwen3.8-Max and plans to release its weights next week, but its multi-day agent results remain...
By Wei-Lin Zhao 5 days agoGraphRAG vs. vector RAG: Match retrieval architecture to the question
Benchmarks favor graphs for multi-hop and corpus-wide questions, while vector RAG remains competitive for direct fact lookup.
By Renata Fuchs 6 days ago
OpenAI Presence brings managed AI agents to enterprise workflows
OpenAI’s limited-GA Presence pairs agents with deployment engineers, controls and testing, while pricing and service terms remain...
By Colin Brandt 6 days ago